자율주행 및 객체 탐지 기반의 실내 보안 로봇 시스템 설계 = Design of an Indoor Security Robot System based on Autonomous Navigation and Object Detection

다국어 초록 (Multilingual Abstract)

In this paper, we propose a robotic system for performing security missions in indoor environments. The proposed system consists of a differential-drive mobile robot platform, a LiDAR, and a camera. The LiDAR is used for map generation, localization, and obstacle avoidance, while the camera is used for human detection and QR code recognition. Human detection is performed using the YOLOv4 object detection algorithm and we have refined this process to improve accuracy. The system is implemented using the Robot Operating System (ROS) 2. To autonomously perform security missions in designated indoor areas, the system generates a map and sets waypoints for autonomous driving. The proposed system requests QR code checks to detected humans by voice to identify users. We have demonstrated the effectiveness of the proposed system through real experiments.

참고문헌 (Reference)

1 A. Bochkovskiy, "Yolov4: Optimal speed and accuracy of object detection", arXiv preprint arXiv:2004.10934"

2 S. Macenski, "The marathon 2 : A navigation system" 2718-2725, 2020

3 D. Fox, "The dynamic window approach to collision avoidance" 4 (4): 23-33, 1997

4 A. Lopez, "Robotman : A security robot for human-robot interaction" 7-12, 2017

5 S. Macenski, "Robot Operating System 2 : Design, architecture, and uses in the wild" 7 : 66-, 2022

6 O. Esan, "Review on Autonomous Indoor Wheel Mobile Robot Navigation Systems" 1-6, 2020

7 김의선 ; 김범수 ; 김익상, "ROS를 활용한 서빙 이동로봇의 구현" 한국정보기술학회 17 (17): 33-43, 2019

8 J. Yu, "Design and implementation of security robot system based on YOLOv3" 2019

9 T. Dharmasena, "Design and Implementation of an Autonomous Indoor Surveillance Robot based on Raspberry Pi" 244-248, 2019

10 I. Chio, "Design and Autonomous Navigation of a New Indoor Disinfection Robot Based on Disinfection Modeling" 20 (20): 649-661, 2023

11 D. Fox, "Adapting the sample size in particle filters through KLD-sampling" 22 (22): 985-1003, 2003